Electrical computers and digital data processing systems: input/ – Intrasystem connection – Bus access regulation
Reexamination Certificate
1999-04-13
2001-10-30
Auve, Glenn A. (Department: 2181)
Electrical computers and digital data processing systems: input/
Intrasystem connection
Bus access regulation
C710S066000
Reexamination Certificate
active
06311248
ABSTRACT:
TECHNICAL FIELD
The present invention pertains to the field of computer system bus architectures. More particularly, the present invention relates to a method and system for optimizing data transfers in a mixed 64-bit/32-bit PCI environment.
BACKGROUND ART
A bus architecture of a computer system conveys much of the information and signals involved in the computer system's operation. In a typical computer system, one or more busses are used to connect a central processing unit (CPU) to a memory and to input/output elements so that data and control signals can be readily transmitted between these different components. When the computer system executes its programming, it is imperative that data and information flow as fast as possible in order to make the computer as responsive as possible to the user. With many peripheral devices, such as graphics adapters, full motion video adapters, small computer systems interface (SICS) host bus adapters, and the like, it is imperative that large block data transfers be accomplished expeditiously. These applications are just some examples of subsystems that benefit substantially from a very fast bus transfer rate.
Much of a computer system's functionality and usefulness to a user is derived from the functionality of the peripheral devices. For example, the speed and responsiveness of the graphics adapter is a major factor in a computer system's usefulness as an entertainment device. Or, for example, the speed with which video files can be retrieved from a hard drive and played by the graphics adapter determines the computer system's usefulness as a training aid. Hence, the rate at which data can be transferred among the various peripheral devices often determines whether the computer system is suited for a particular purpose. The electronics industry has, over time, developed several types of bus architectures. Recently, the PCI (peripheral component interconnect) bus architecture has become one of the most widely used, widely supported bus architectures in the industry. The PCI bus was developed to provide a high speed, low latency bus architecture from which a large variety of systems could be developed.
Prior Art
FIG. 1
shows a typical PCI bus architecture
100
. PCI bus architecture
100
is comprised of a CPU
102
and a main memory
104
, coupled to a host PCI bridge containing arbiter
106
(hereafter arbiter
106
) through a CPU local bus
108
and memory bus
110
, respectively. A PCI bus
112
is coupled to each of PCI agents
114
,
116
,
118
,
120
,
122
,
124
respectively, and is coupled to arbiter
106
.
Referring still to Prior Art
FIG. 1
, each of PCI agents
114
,
116
,
118
,
120
,
122
,
124
(hereafter, PCI agents
114
-
124
) residing on PCI bus
112
use PCI bus
112
to transmit and receive data. PCI bus
112
is comprised of functional signal lines, for example, interface control lines, address/data lines, error signal lines, and the like. Each of PCI agents
114
-
124
are coupled to the functional signal lines comprising PCI bus
112
. When one of PCI agents
114
-
124
requires the use of PCI bus
112
to transmit data, it requests PCI bus ownership from arbiter
106
. The PCI agent requesting ownership is referred to as an “initiator”, or bus master. Upon being granted ownership of PCI bus
112
from arbiter
106
, the initiator (e.g., PCI agent
116
) carries out its respective data transfer.
Each of PCI agents
114
-
124
may independently request PCI bus ownership. Thus, at any given time, several of PCI agents
114
-
124
may be requesting PCI bus ownership simultaneously. Where there are simultaneous requests for PCI bus ownership, arbiter
106
arbitrates between requesting PCI agents to determine which requesting PCI agent is granted PCI bus ownership. When one of PCI agents
114
-
124
is granted PCI bus ownership, it initiates it transaction (e.g., data transfer) with a “target ” or slave device (e.g., main memory
104
). When the data transaction is complete, the PCI agent relinquishes ownership of the PCI bus, allowing arbiter
106
to reassign PCI bus
112
to another requesting PCI agent.
Thus, only one data transaction can take place on a PCI bus at any given time. In order to maximize the efficiency and data transfer bandwidth of PCI bus
112
, PCI agents
114
-
124
follow a definitive set of protocols and rules. These protocols are designed to standardize the method of accessing, utilizing, and relinquishing PCI bus
112
, so as to maximize its data transfer bandwidth. The PCI bus protocols and specifications are set forth in an industry standard PCI specification (e.g., PCI Specification—Revision 2.1). Where each of PCI agents
114
-
124
are high performance, well designed devices, data transfer rates of up to 528 Mbytes per second can be achieved (e.g., PCI bus
112
operating at 66 MHz).
Prior art
FIG. 2
shows a diagram of arbiter
106
, PCI bus
112
, PCI target
114
, and PCI target
116
in greater detail. As depicted in
FIG. 2
, PCI bus
112
is a 64-bit PCI bus (in accordance with PCI specification—Revision 2.1). However, PCI bus
112
hosts a mixed 32-bit and 64-bit PCI environment. PCI target
114
is a 64-bit PCI device, while PCI target
116
is a 32-bit PCI device. In accordance with the PCI specification, a 64-bit PCI bus (e.g., 64-bit PCI bus
112
) must be fully functional and adequately host both 32-bit and 64-bit PCI devices. With both 32-bit and 64-bit devices coupled to a common 64-bit PCI bus (e.g., 64-bit PCI bus
112
), a 64-bit PCI initiator, when it starts a cycle, is required to determine the width of the target (e.g., whether the target device is a 32-bit or 64-bit device), by asserting the REQ64# signal and waiting for the ACK64# signal being asserted with DEVSEL# by the target device. If the target device asserts ACK64# with DEVSEL#, then the target device is notifying the initiator that it is a 64-bit device. If ACK64# is not asserted with DEVSEL#, then the 64-bit initiator determines that the target device is a 32-bit device.
Thus, once the 64-bit PCI initiator
106
determines that the target device
116
is a 32-bit PCI target, the 64-bit PCI initiator
106
performs “data-steering” to transfer the data from its data transaction 32-bits at a time. However, there exists a problem in the fact that single data transaction 64-bit transfers are very inefficient using this scheme.
As is well known, using a single data phase with 64-bit transfers may not be very effective. Since the 64-bit PCI initiator
106
does not know how the transaction will be resolved with ACK64# until DEVSEL# is returned, it does not know the clock on which to deasseirt FRAME# for a 64-bit single dataphase transaction. IRDY# must remain deasserted, until FRAMES signaling is resolved. The single 64-bit data phase may have to be split into two 32-bit data phases when the target is only 32-bits, which means a two phase 32-bit transfer is a least as fast as a one phase 64-bit transfer.
Prior art
FIG. 3
shows a timing diagram
300
of 64-bit PCI initiator
106
transferring a 64-bit data, using the REQ64#, ACK64# protocol, to 64-bit PCI target
114
. The left side of timing diagram
300
shows the standard signals of 64-bit PCI bus
112
. The letter “Z” signifies that the relevant signal lines are tri-stated. As shown by arrows
301
, it should be noted that until ACK64# is sampled asserted by 64-bit PCI initiator
106
, FRAME# cannot be deasserted and IRDY# cannot be asserted, which causes a 3-clock cycle data transfer transaction, when there is only one 64-bit data to transfer.
Prior art
FIG. 4
shows a timing diagram
400
of 64-bit PCI initiator
106
transferring a 64-bit data, using the REQ64#, ACK64# protocol, to a 32-bit PCI target
116
. Diagram
400
shows a case where 64-bit PCI initiator
106
transfers a 64-bit data to 32-bit PCI target
116
. As shown in diagram
400
, when 64-bit PCI initiator
106
samples that the ACK64# is not asserted with DEVSEL#, the initiator kno
Chambers Peter
Meiyappan Subramanian S.
Auve Glenn A.
VLSI Technology Inc.
Wagner , Murabito & Hao LLP
LandOfFree
Method and system for optimized data transfers in a mixed... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for optimized data transfers in a mixed..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for optimized data transfers in a mixed... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2595303